Using RelocInfo instead of DebuggerStatementStub

Committed: http://code.google.com/p/v8/source/detail?r=3814

[Søren Thygesen Gjesse, 2010-02-03 21:46:45]

LGTM

http://codereview.chromium.org/561049/diff/1/24
File src/heap.h (right):

http://codereview.chromium.org/561049/diff/1/24#newcode104
src/heap.h:104: V(Code, debugger_statement_code, DebuggerStatementCode)         
            \
You should be able to remove this root as well.

[SeRya, 2010-02-04 14:29:21]

Removing debug_statement_code I have to fix the ExitFrame::code method with uses
it.

A I understand ExitFrame::code and MacroAssembler::EnterExitFramePrologue (or
MacroAssembler::EnterExitFrame in ARM) both use zero to mark the debug C-entry
stub. Since this mark is not used anywhere else I left only general case (at
least I haven't found any using and removing it doesn't bake tests).

The changes code was introduced in this CL:
http://codereview.chromium.org/330017 which was reviewed by Erik Corry. Erik,
please, review my CL too (at least the
ExitFrame::code/MacroAssembler::EnterExitFramePrologue part).

http://codereview.chromium.org/561049/diff/1/24
File src/heap.h (right):

http://codereview.chromium.org/561049/diff/1/24#newcode104
src/heap.h:104: V(Code, debugger_statement_code, DebuggerStatementCode)         
            \
On 2010/02/03 21:46:47, Søren Gjesse wrote:
> You should be able to remove this root as well.

Done.

[Søren Thygesen Gjesse, 2010-02-05 14:03:34]

LGTM

I think storing 0 into the code slot is a leftover which is not used any more.

Lasse, I seem to remember that you removed some of the special markers stored in
the code slot at some point (during the x64 port perhaps). Was this maybe left
at that point?

http://codereview.chromium.org/561049/diff/4027/3121
File src/arm/macro-assembler-arm.cc (right):

http://codereview.chromium.org/561049/diff/4027/3121#newcode337
src/arm/macro-assembler-arm.cc:337: push(ip);  // accessed from
EditFrame::code_slot
Initial letter uppercase, end with period, no class called EditFrame.

http://codereview.chromium.org/561049/diff/4027/3116
File src/assembler.cc (right):

http://codereview.chromium.org/561049/diff/4027/3116#newcode432
src/assembler.cc:432: return "code target (context)";
Change this to

    case RelocInfo::DEBUG_BREAK:
#ifdef ENABLE_DEBUGGER_SUPPORT
      return "debug break";
#else
      UNREACHABLE();
#endif

http://codereview.chromium.org/561049/diff/4027/3116#newcode493
src/assembler.cc:493: case CODE_TARGET_CONTEXT:
Something like above, or just remove the #ifdef/#endif.

http://codereview.chromium.org/561049/diff/4027/3108
File src/assembler.h (right):

http://codereview.chromium.org/561049/diff/4027/3108#newcode122
src/assembler.h:122: #ifdef ENABLE_DEBUGGER_SUPPORT
I think we should avoid the #ifdef/#endif here to always have the same id for
reloc modes.

http://codereview.chromium.org/561049/diff/4027/3102
File src/ia32/macro-assembler-ia32.cc (right):

http://codereview.chromium.org/561049/diff/4027/3102#newcode407
src/ia32/macro-assembler-ia32.cc:407: push(Immediate(CodeObject()));  //
accessed from EditFrame::code_slot
See previous comment on comment.

http://codereview.chromium.org/561049/diff/4027/3110
File src/x64/macro-assembler-x64.cc (right):

http://codereview.chromium.org/561049/diff/4027/3110#newcode1965
src/x64/macro-assembler-x64.cc:1965: push(kScratchRegister);  // accessed from
EditFrame::code_slot
See previous comment on comment.

[SeRya, 2010-02-05 15:47:16]

http://codereview.chromium.org/561049/diff/4027/3121
File src/arm/macro-assembler-arm.cc (right):

http://codereview.chromium.org/561049/diff/4027/3121#newcode337
src/arm/macro-assembler-arm.cc:337: push(ip);  // accessed from
EditFrame::code_slot
On 2010/02/05 14:03:34, Søren Gjesse wrote:
> Initial letter uppercase, end with period, no class called EditFrame.

Done.

http://codereview.chromium.org/561049/diff/4027/3116
File src/assembler.cc (right):

http://codereview.chromium.org/561049/diff/4027/3116#newcode432
src/assembler.cc:432: return "code target (context)";
On 2010/02/05 14:03:34, Søren Gjesse wrote:
> Change this to
> 
>     case RelocInfo::DEBUG_BREAK:
> #ifdef ENABLE_DEBUGGER_SUPPORT
>       return "debug break";
> #else
>       UNREACHABLE();
> #endif

at the case of  RelocInfo::NUMBER_OF_MODES both are used (UNREACHABLE and
return) so I do the same.

http://codereview.chromium.org/561049/diff/4027/3116#newcode493
src/assembler.cc:493: case CODE_TARGET_CONTEXT:
On 2010/02/05 14:03:34, Søren Gjesse wrote:
> Something like above, or just remove the #ifdef/#endif.

Done.

http://codereview.chromium.org/561049/diff/4027/3108
File src/assembler.h (right):

http://codereview.chromium.org/561049/diff/4027/3108#newcode122
src/assembler.h:122: #ifdef ENABLE_DEBUGGER_SUPPORT
On 2010/02/05 14:03:34, Søren Gjesse wrote:
> I think we should avoid the #ifdef/#endif here to always have the same id for
> reloc modes.

Done.

http://codereview.chromium.org/561049/diff/4027/3102
File src/ia32/macro-assembler-ia32.cc (right):

http://codereview.chromium.org/561049/diff/4027/3102#newcode407
src/ia32/macro-assembler-ia32.cc:407: push(Immediate(CodeObject()));  //
accessed from EditFrame::code_slot
On 2010/02/05 14:03:34, Søren Gjesse wrote:
> See previous comment on comment.

Done.

http://codereview.chromium.org/561049/diff/4027/3110
File src/x64/macro-assembler-x64.cc (right):

http://codereview.chromium.org/561049/diff/4027/3110#newcode1965
src/x64/macro-assembler-x64.cc:1965: push(kScratchRegister);  // accessed from
EditFrame::code_slot
On 2010/02/05 14:03:34, Søren Gjesse wrote:
> See previous comment on comment.

Done.